Missing Ice Candidate between v1.11.0 and v1.11.1

[SixK]

Which version are you using?

v1.11.1

Which operating system are you using?

Linux amd64 standard

Describe how to replicate the issue

I notice a difference between v1.11.1 and v1.11.0 in ice candidate gathering.
On our platform the following ice candidate does not exists anymore when looking in about:webrtc firefox page.

| succeeded | true | true | (redacted):5019/tcp(prflx) [non-proxied] | 95.64.103.123:8190/tcp(host) |

This ice candidate is probably related to webrtcAdditionalHosts public IP we have added in mediamtx.yml.
Sadly this is the only ice candidate pair working in our restricted network when connected to our corporate VPN.
If directly connected to internet then UDP candidates are working.

If blocking UDP in mediamtx modifying webrtcLocalUDPAddress parameter,
then no stream is working anymore regardless network we are using.

Our infrastructure is complex with public IP, load-balancers, virtual machines, restricting network rules, etc...

So I guess moving from pions/ice v2 to v4 (or pions/webrtc v2 to v4 or maybe a recent fix) has broken something in TCP candidate gathering ?
But I have no clue of what is wrong.
Comparing logs of our involved applications I see no real difference.

This issue is related to this discussion #4153

Server logs

Sorry no logs for security and I found nothing relevant anyway even in debug mode comparing v1.11.0 and v1.11.1 outputs line by line

Network dump

No response

[SixK]

Doing some extra test, I changed webrtcIPsFromInterfaces from no to yes in mediamtx.yml and missing ice candidate is back ! :)
So bug seem's related to webrtcAdditionalHosts and webrtcIPsFromInterfaces combination.

In v1.11.0:
webrtcAdditionalHosts: [102.34.23.56]
webrtcIPsFromInterfaces: no
is working

In v1.11.1 I need to use:
webrtcAdditionalHosts: [102.34.23.56]
webrtcIPsFromInterfaces: yes

I this case, I get useless extra local interfaces candidates, but my missing candidate is back.